Yes..that makes sense.

Not at all surprising, since the code is converted to c++, and then
compiled.

if you every tried to run null loops in c to create a delay..you in for a
surprise..as the compiler will completely skip the code...

for x = 1; x <= 100000; x++)
{};


In VB6, VBA, or mvBasic, those loops are executed...even when they do
nothing.

However, in c, the above loop is looked at by the compiler...and simply
skipped.....

The same thing is done for redundant variable assigns etc. so, even if you
go:

for x = 1; x <= 100000; x++)
{y = 123};

Even in the above, the compiler figures out no other variables and refs
occur to "y", so it moves it out of the loop......once again, the compiler
will simply skip the above loop....

So, not one bit surprised that jBase skips the loop, since jBase code is
converted to c..and then compiled....

--
Albert D. Kallal
Edmonton, Alberta Canada
pleaseNOOSpamKallal (AT) msn (DOT) com

What C compiler do:

int i=10, j=11;
for( i = 1, j= 0; i<10 ; i++)
{
....
}

xxxxx

1 in any case sets i to 1
2 in any case sets j to 0
3 checks condition i < 10 if TRUE then goto4 else goto7
4 executes {...}
5 executes i++
6 goto3
7 xxxxx


In case: FOR I=10 TO 1
I think that DataBasic compiler checks that condition is constant and
always FALSE. In this case compiler ignores entire FOR.
Regards,
Grigory

I can picture the code that's being executed in the 10 TO 1 issue.
It's always been my understanding that the FOR statement checks to see
if the incremented value exceeds the limit, then does the increment
and enters the loop if it does not. It seems most platforms assign
the initial value, then check, where Universe checks then increments.
We're talking about the placement of one high level statement (2-3
assembler lines?). In another life that wouldn't be wrong but at this
point if they want to follow established precedent then it is wrong.

T

John Henderson wrote:

Tony Gravagno wrote:

Precedent aside, and having had a look at the Basic Commands
Reference, I agree. The behaviour I see is wrong. For the
statement

FOR variable = start TO end [STEP increment]

the manual says (without reservation) "variable is assigned the
value of start, which is the initial value of the counter."

QED.

John

In an ideal world, depending on the index variable outside the scope of the
For-Next is not allowed--fortran et al leave it undefined. But I've written
many times (before we had EXIT):

FOR i= 1 to imax
IF (terminating condition) THEN i=999999
END ELSE
END
NEXT i
IF i >= 999999 THEN
error
END

I could've, of course set imax to -1 which is a better test anyway.

Chandru Murthi

"John Henderson" <jhenRemoveThis (AT) talk21 (DOT) com> wrote

On Feb 11, 2:06 pm, Tony Gravagno
<g6q3x9lu53... (AT) sneakemail (DOT) com.invalid> wrote:
NO, this has nothing to do with that. BASIC does not have any scoping.
It is to do with whether the compiler decides it can do any
opimizations of loop initialization vs iteration vs having the loop at
all. I suspect that the UniVerse compiler here is incorrect and that
the variable A should be intialized to 1 even if the loop code is
eliminated.

I don't remember making the jBASE compiler do anything that would
cause the result to be 9, so if this is truly the resutl on jBASE 4,
then I suspect that someone played with the compiler or runtime for
loops at the jBASE 4.0 level and it has been corrected at 4.1. Someone
should check jBASE 3.x although it is farily irrelevant as the code is
in error of course and one should not program for side effects (thoguh
perhaps people do not see that they are doing so if this was FOR X=A
TO B STEP V for instance.

However I DO vaguely remember making the compiler check that the start
and end points made sense when the arguments were literal constants.
It might be that this warning would only be issued if you used (W4
warning levels on the jBASE compiler.

The reason for the warning is because the behavior of syntactically
sound but semantically incorrect code is undefined. This is the case
in many languages and compilers. In the case of jBASE I was not
exactly hampered by but had to pay attention to the behavior of
existing systems. Sounds like this particular case was overlooked at
some point in development.

Other 'interesting' cases you might like to try if you are bored, are:

IF 3=4 THEN PRINT 5
IF 4=4 THEN PRINT 6

Jim

"Jim Idle" <jimi (AT) temporal-wave (DOT) com> wrote

code such as:

if FLAG=FALSE then FLAG=FALSE

I'd say Why? And he's say he just was making sure. He didn't trust the
computer.


Mark Brown

On a related note:
Since line 1 in this code is mandatory:
0001 STRING = ''
0002 STRING<I> = some.value

I see code all the time where
0001 STRING = ''
0002 STRING = some.value

Jes' makin' sure, eh?

Chandru

"Mark Brown" <mbrown (AT) drexelmgt (DOT) com> wrote

On Feb 13, 9:12 am, "Chandru Murthi"
<cmur_xyz_thi (AT) xyz_seeinggree_xyz_n (DOT) net> wrote:
That is incorrect regarding Fortran. For the code

do i=1,2
print*,i
end do
print*,i

the variable "i" has value 3 after exiting the loop, and the code must
give the result

1
2
3

Last I used Fortran was in 1971 on an IBM 370, so quite possibly things have
changed since then. My point was only that it may not be good programming
practice to depend on an index var's value, it's better to use your own
flags.

Chandru

"Beliavsky" <beliavsky (AT) aol (DOT) com> wrote